A new framework of biotechnology benchmarks aims to determine if AI models possess the capability to solve the most complex problems in biology. Edison Scientific and FutureHouse assembled these challenges, known as the Millennium Problems for Biology, to be difficult to solve but easy to validate in a simple laboratory environment. Solutions to these problems would provide material contributions toward curing diseases.

The initiative responds to the lack of rapid validation tools for AI-driven drug discovery, contrasting with the immediate verifiability of mathematical breakthroughs like the Navier-Stokes problem. FutureHouse and Edison Scientific are currently recruiting community members to red-team the specifications to ensure they are difficult to cheat. The final list will be updated based on this feedback and hosted on the Bio Millennium Problems website.